As their prices decline, their storage capacities increase, and their endurance improves, NAND Flash Solid-State Disks (SSD) provide an increasingly attractive alternative to Hard Disk Drives (HDD) for portable computing systems and PCs. HDDs have been an integral component of computing systems for several decades as long-term, non-volatile storage in memory hierarchy. Today's typical hard disk drive is a highly complex electro-mechanical system which is a result of decades of research, development, and fine-tuned engineering. Compared to HDD, flash memory provides a simpler interface, one without the complexities of mechanical parts. On the other hand, today's typical solid-state disk drive is still a complex storage system with its own peculiarities and system problems.

Due to lack of publicly available SSD models, we have developed our NAND flash SSD models and integrated them into DiskSim, which is extensively used in academe in studying storage system architectures. With our flash memory simulator, we model various solid-state disk architectures for a typical portable computing environment, quantify their performance under real user PC workloads and explore potential for further improvements. We find the following:

• The real limitation to NAND flash memory performance is not its low per-device bandwidth but its internal core interface.

• NAND flash memory media transfer rates do not need to scale up to those of HDDs for good performance.

• SSD organizations that exploit concurrency at both the system and device level improve performance significantly.

• These system- and device-level concurrency mechanisms are, to a significant degree, orthogonal: that is, the performance increase due to one does not come at the expense of the other, as each exploits a different facet of concurrency exhibited within the PC workload.

• SSD performance can be further improved by implementing flash-oriented queuing algorithms, access reordering, and bus ordering algorithms which exploit the flash memory interface and its timing differences between read and write requests.